Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
FS#18759 - [initscripts] Obsolete escape sequence aborts backgrounded daemons
Attached to Project:
Arch Linux
Opened by kujub (kujub) - Saturday, 20 March 2010, 12:58 GMT
Last edited by Roman Kyrylych (Romashka) - Tuesday, 04 May 2010, 09:48 GMT
Opened by kujub (kujub) - Saturday, 20 March 2010, 12:58 GMT
Last edited by Roman Kyrylych (Romashka) - Tuesday, 04 May 2010, 09:48 GMT
|
DetailsDescription:
When backgrounding slow custom scripts in DAEMONS, these are silently aborted as soon as the set_consolefont function gets called. (Bash shell exit without sigterm verified with 'trap'.) This happens with both fbsplash-scripts and initscripts-extras-fbsplash which need to defer 'setfont' to the end of rc.multi. I was able to track this down to printf '\033(K' which is the escape sequence to make console maps active. According to the man pages this sequence seems to be only needed to be sent by the scripts when using 'mapscrn' whereas 'setfont' sends it by itself. Therefore the printf if obsolete and should be dropped AFAIKS. Stuff works as normal here without this. Additional info: * package version(s) initscripts git * config and/or log files etc. /etc/rc.conf CONSOLEFONT=ter-v16n CONSOLEMAP=cp850 # just for testing Steps to reproduce: Use initscripts hooks deferring setfont to rc.multi by unsetting CONSOLEFONT in rc.sysinit and calling set_consolefont from multi_end hook together with some slow script backgrounded in DAEMONS. 
0001-Remove-obsolete-consolem...
(0.8 KiB)
|
This task depends upon
Closed by Roman Kyrylych (Romashka)
Tuesday, 04 May 2010, 09:48 GMT
Reason for closing: Fixed
Additional comments about closing: applied the correcting patch to git
Tuesday, 04 May 2010, 09:48 GMT
Reason for closing: Fixed
Additional comments about closing: applied the correcting patch to git
the patch breaks the current behaviour in the following scenario:
1) let the system be configured in a non-UTF8 mode and with consolemap
2) reset the terminal (either run `reset` or use any broken app that does it for you)
3) now you have a broken console (this is only the case for non-UTF8 mode, because in recent kernels UTF8 is the default)
4) exit and login back on the same console
now, \e%@ will be executed, because it is written in locale.sh, but \e(K won't (because the patch removed it)
- the console will still be broken, while before the patch it would be restored back to the configured state
I think that removing the 'for' loop is enough to solve the problem you described, but please keep
echo 'if [ "$CONSOLE" = "" -a "$TERM" = "linux" -a -t 1 ]; then printf "\033(K"; fi' >>/etc/profile.d/locale.sh
P.S.: whether locale.sh should restore the broken-after-reset console is another question.
I think it's convenient and does not hurt, but if people think that it should be removed - then the same should be done for other magic echos that are written to locale.sh, because they are only really needed for restoring the console after a reset.